License Plate Recognition systems (LPR), also called Automatic Number Plate Recognition systems (ANPR), take an important role in modern traffic control systems
Abstract
License Plate Recognition systems (LPR), also called Automatic Number Plate Recognition systems (ANPR), take an important role in modern traffic control systems (law enforcement, toll collection, surveillance etc.). Typically, such systems operate under restricted conditions, such as fixed illumination and stationary background, most systems use special cameras, sometimes a set of several cameras with built-in illumination. Also, such system must usually be installed and calibrated depended on its location and purpose.
In this work we propose a new algorithm for a different type of license plate recognition system, where a simple (analogue, monochrome) video camera is used, without controlling the lighting conditions and without need of adjusting the system to a specific site of installation.
Our algorithm searches a video input from a simple camera for Israeli license plates, and cuts them out of the frame, separates the numbers from each other which enables then to use a standard character recognition system (OCR) for identification.
The algorithm is based on a combination of different approaches, including motion detection, analysis of edge statistics, typical pattern recognition, geometric filtering, and a learning system for final classification of candidate license plate objects.
The algorithm was tested in MATLAB environment on video samples from a surveillance camera of a potential customer. Our results show a very good detection rate and low false detection rate and run in short time. Moreover license plates are consistently found in concurrent frames.
Flowchart
Results
